Download websites that use particles.js

particles.js draws a field of moving dots on a canvas element, joins nearby dots with thin lines, and reacts to the cursor. Vincent Garreau published it as a single JavaScript file driven by a large configuration object, under MIT and dated 2014 in our record, and the effect quickly became the default way to make a dark landing page look technical. Startups, crypto projects, agencies and conference sites reached for it in the same handful of years, which is why the library reads today as a date stamp on a design.

As of November 2026 we detect particles.js on 476,934 sites. Very few of those sites depend on the library for anything functional, because the effect is decorative by definition and a page renders correctly without it. What the presence of the file signals instead is a design decision, and usually a hero section built during a particular stretch of front end fashion.

There is a technical implication as well. A full screen canvas animation runs continuously, keeps the main thread busy and drains phone batteries, so the library surfaces in performance audits as a Core Web Vitals problem on sites whose owners have no idea that a background is costing them. Anyone assembling a sample of pages with avoidable client side animation, or studying how decorative code correlates with poor field performance, has a clean population here, because the fingerprint is unambiguous when the file loads.

The footprint is flat rather than clustered. United States accounts for 138,310 of the total and .com leads the domain mix, but the remainder is spread across markets in a way that regionally sold products are not, which fits a free file copied out of a public repository by developers everywhere. The original repository stopped receiving updates long ago and a rewritten successor exists under a different name, so a site loading the original is running unmaintained code, though unmaintained decoration carries far less risk than an unmaintained form handler or payment script.

One limit is worth naming plainly. Detection sees the script, not its settings, so the dataset cannot tell you whether a page draws a dense mesh or a few slow dots, and a bundled build that inlines the library into an application file escapes the fingerprint entirely, which makes the count a floor. Within that limit, a per-domain check answers only whether one site loads the file, while the downloadable list gives the population: which domains, which countries, which co-loaded frameworks, and how the set shrinks as sites are rebuilt, which is the measurement anyone tracking front end churn actually wants.